Re: [PATCH] audit: add feature audit_lost reset

[Steve Grubb <sgrubb@redhat.com>]

On Wednesday, January 11, 2017 11:19:42 PM EST Richard Guy Briggs wrote:
> > OK. the code to support this is in svn. However, since we didn't use a
> > feature bit like we normally do, there is absolutely no way to report
> > that the underlying kernel does not support this. It quietly fails and
> > pretends everything is fine. I'd prefer that we had a feature bit to
> > output a proper error message.
> 
> Do you still want to switch to CONFIG_CHANGE?  (I think that is a good
> idea.)

Sure.

> I agree detecting this feature is a destructive operation requiring an
> existing lost count and checking the positive return code, but not
> impossible, and would prefer a feature bit.

I'd prefer a feature bit so that I can tell people your kernel doesn't support 
this. Audit runs on a large variety of kernels.

> As for audit being immutable, I could see an argument to have this
> feature usable even though the config is locked.  What's your take?

I can see value in resetting the count even when immutable. Perhaps just use 
its logging function. So we don't have a new record type.

-Steve